Building a Network Knowledge Base Based on a Belief Revision Operator.

We present a proposal for negotiation in a network knowledge base ( NKB ) formed by a set of k intelligent agents, in order to decide the acceptance (or not) of a new information ϕ that is expressed in propositional logic. Each one of the agents Ai , i = 1,…, k has associated a knowledge base ( KB ) Ki expressed as propositional formulas in conjunctive form. In order to decide the acceptance of ϕ , the belief revision operator Ind is applied on each KB Ki and ϕ , then Si = Ind ( ϕ , Ki ), and each one of the literals that form Si is weighed in order to estimate the cost that is required by Ai to accept ϕ . Si expresses the required information by the agent Ai such that ( Ki ∪ Si ) ╞ ϕ . The NKB uses a constant value γ that is the threshold of acceptance of NKB . Thus, a piece of new information ϕ will be accepted by all the agents in NKB if the average of the weights of the requested formulas Si is greater than γ . Otherwise, ϕ is rejected by all the agents in the NKB . Only in the case that ϕ will be accepted collectively, each one of the Ki in the network NKB is updated, as Ki = ( Ki ⋀ Si ). Otherwise, there are no changes to any KB Ki in the NKB .